Skip to Content
View all events

Migrating Spec-less Code with AI and Parity

Published by O'Reilly Media, Inc.

Intermediate content levelIntermediate

From naive code translation to absolute behavioral parity

What you’ll learn and how you can apply it

  • Analyze the systemic failure modes of naive AI translation
  • Architect a black-box parity testing framework
  • Enforce absolute functional equivalence between disparate language ecosystems

Course description

Blindly relying on AI to translate legacy source code from one language or framework to another, such as converting Java to Python via simplistic prompts like “translate this folder containing a Java application to a new Python application” is guaranteed to lead to catastrophic production failures. Standard AI coding tools lack the architectural awareness to capture hidden dependencies, language-specific runtime nuances, and edge cases embedded within undocumented legacy systems.

Fernando Vieira shows you why naive code conversion inevitably fails and introduces a rigorous, engineering-first framework designed to eliminate modernization risk entirely. Instead of guessing your way through a migration, you’ll discover how to establish black-box parity as an automated quality gate between your old and new systems. Through a practical, live cross-language showcase, you’ll see how to capture real-world operational behaviors from spec-less legacy code and use them to construct a bulletproof functional contract. You’ll take away a predictable and repeatable protocol to enforce absolute behavioral equivalence, ensuring your modernized application maintains the exact execution dynamics of the original system without relying on nonexistent documentation.

This live event is for you because...

  • You’re a software professional tasked with migrating, refactoring, or modernizing legacy code bases and you need a reliable, deterministic process to execute cross-technology transitions safely.
  • You’re engaged in legacy system modernization and have experienced the limitations, regressions, or silent failures that occur when relying on standard AI code-generation assistants for blind language-to-language translation.

Prerequisites

  • Intermediate to advanced software development skills
  • Basic experience using AI coding assistants for development tasks

Recommended preparation:

Recommended follow-up:

Schedule

The time frames are only estimates and may vary according to how the class is progressing.

The blind translation trap (20 minutes)

  • Poll: Have you tried using AI to convert code from language A to B?
  • Presentation: Why blind syntax translation fails; shifting the paradigm to black-box parity

Code conversion under a parity loop (30 minutes)

  • Presentation and demonstration: Presenting an undocumented legacy component; showing the set of sample production inputs and outputs used as the invariant baseline; launching the multi-agent execution context where a generation agent writes the target code and a validation agent instantly executes the parity gate—feeding inputs, comparing outputs against the baseline, and forcing self-correction until absolute equivalence is reached

Wrap-up and Q&A (10 minutes)

  • Presentation: What you should remember from this protocol

Your Instructor

Fernando J. Vieira

Fernando J. Vieira is a Brazilian-Portuguese software craftsman with over two decades of industry experience. His current focus at Thoughtworks is on cloud modernization projects. Bringing a global and multidisciplinary perspective, Fernando holds a robust academic foundation in software engineering, electrical engineering, and data science. His extensive background spans multiple continents and sectors ranging from traditional manufacturing to fast-paced tech startups. Grounded in the principles of software craftsmanship, Fernando partners with teams to accelerate the adoption of modern practices and elevate engineering discipline. He delivers high-impact results through hands-on mentoring, targeted instruction, and direct engagement across development practices, CI/CD, automation, and infrastructure as code. Fernando is also an active voice in the tech community, contributing through articles on software excellence and

Skill covered

Refactoring